Jump to main content.


Utility of Three Types of Mass Spectrometers for Determining Elemental Compositions of Ions Formed from Chromatographically Separated Compounds

Andrew H. Grange,1 Floyd A. Genicola,2 and G. Wayne Sovocool 1
1Environmental Sciences Division, NERL,
U.S. EPA, P.O. Box 93478,
Las Vegas, NV 89193-3478
2Office of Coastal Planning & Program Coordination
N.J. Department of Environmental Protection
PO Box 418, 401 E. State Street, Trenton, NJ 08625-0418

ABSTRACT 

Concentration factors of 1000 and more reveal dozens of compounds in extracts of water supplies.&nbsp Library mass spectra for many of these compounds are not available, and alternative means of identification are needed.&nbsp Determination of the elemental compositions of the ions in mass spectra makes feasible searches of commercial and chemical literature that often lead to compound identification.&nbsp Instrumental capabilities that constrain the utility of a mass spectrometer for determining ion compositions for compounds that elute from a chromatographic column are scan speed, mass accuracy, linear dynamic range, and resolving power.&nbsp Mass Peak Profiling from Selected Ion Recording Data (MPPSIRD) performed with a double focusing mass spectrometer provides the best combination of these capabilities.&nbsp This technique provides unique ion compositions for ions of higher mass from compounds eluting from a gas chromatograph than can be obtained by orthogonal acceleration time-of-flight (oa-TOF) or Fourier transform ion cyclotron resonance mass spectrometry.&nbsp Multiple compositions are usually possible for an ion with a mass exceeding 150 Da within the error limits of the mass measurement.&nbsp The correct composition is selected based on measured exact masses of the mass peak profiles resulting from isotopic ions higher in mass by 1 and 2 Da and accurate measurement of the summed abundances of these isotopic ions relative to the monoisotopic ion.&nbsp A Profile Generation Model (PGM) automatically determines which compositions are consistent with measured exact masses and relative abundances.&nbsp The utility of oa-TOF and double focusing mass spectrometry using Ion Composition Elucidation (MPPSIRD plus the PGM) are considered for determining ion compositions of two compounds found in drinking water extracts and a third compound from a monitoring well at a landfill.

INTRODUCTION

Concern about endocrine disruption by ultra-trace amounts of water contaminants(1) has encouraged use of 1000-fold and greater(2-4) concentration factors to reveal dozens of unidentified compounds in surface waters, well waters, and sewage treatment plant effluents.&nbsp Only a small fraction of these compounds is targeted by standard analytical methods, found in mass spectral libraries, or has known toxicological effects.&nbsp The EPA lists 2800 chemicals with production levels of at least 106 lbs/year.(5) The EPA's Endocrine Disruptor Screening and Testing Advisory Committee estimates 87,000 compounds are used in commerce.(6) These compounds, byproducts, and degradation products might be found in drinking water sources, air, and contaminated sites.&nbsp Identification of these compounds is necessary to assess risk to humans and aquatic ecosystems.&nbsp Hence, there is a need for more powerful analytical techniques to identify such compounds, which have been neglected due to a lack of generally applicable identification methods.

To limit the need for pre-analysis fractionations, compound identification techniques must isolate signals from low-level contaminants in complex mixtures.&nbsp Even a minor component could pose the greatest health risk due to endocrine disrupting effects or other potent toxicity.&nbsp Excellent component separation is realized by high resolution gas chromatography interfaced to mass spectrometry, a combination that provides separation in time coupled to selection by ion masses.

When no plausible mass spectral library matches are found for an analyte, determination of the elemental compositions of the molecular ion and the fragment ions limits the number of possible compounds sufficiently to make searches of the chemical and commercial literature feasible.&nbsp Subsequent purchase or synthesis of one or a few standards can then confirm tentative identifications.

This article examines four operational characteristics of a mass spectrometer that constrain its utility for determining ion compositions for compounds eluting from a GC or other high resolution separatory technique providing similar chromatographic peak widths: scan speed, the error limits of exact mass measurements, the linear dynamic range of the mass analyzer, and its resolving power when mass interferences are encountered.&nbsp Also discussed is the considerable benefit of measuring exact masses of the isotopic profiles and their relative abundances.

One recent paper compares the utility of orthogonal acceleration time-of-flight (oa-TOF) and Fourier transform ion cyclotron resonance (FTICR) mass spectrometers for determining ion compositions.(7) Here, the strengths and limitations of double focusing, oa-TOF, and FTICR mass spectrometers for this purpose are considered.&nbsp Finally, three real-world examples are illustrated where determination of ion compositions was used to identify, confirm, or characterize compounds found in extracts of well water.

EXPERIMENTAL

Software was developed in-house (U.S. EPA) to plot mass peak profiles from selected-ion-recording data (MPPSIRD) for eluting analytes from which exact masses and relative abundances are determined.&nbsp In addition, a Profile Generation Model (PGM) was written to automatically interpret the data to establish the correct elemental composition.&nbsp Use of MPPSIRD and the PGM in concert is called Ion Composition Elucidation (ICE).(8)

Mass peak profiling from selected ion recording data (MPPSIRD)

Multiple m/z ratios across individual mass peak profiles are monitored using selected ion recording (SIR). The magnetic field is held constant and the linked accelerating potential and electrostatic sector voltage jump among values corresponding to m/z ratios listed by the user in the multiple ion detection (MID) or SIR window.&nbsp For each m/z ratio, the area under the chromatographic peak observed in the ion chromatogram for the corresponding analyte is integrated and plotted to provide a point on a full or partial mass peak profile.&nbsp Figure 1a shows four ion chromatograms corresponding to the maxima of the first four partial profiles in Figure 1b.

Figure 1 - For further information contact grange.andrew@epa.gov

Figure 1 (a) Ion chromatograms acquired with 20,000 resolving power (10% valley) corresponding to the m/z values at the maxima in the first four partial mass peak profiles in (b) including the lock-mass ion at m/z 405.&nbsp The shaded areas are the y-coordinates for these maxima.&nbsp Areas under ion chromatograms for 31 m/z ratios were used to plot partial profiles for each of the two calibration ions (5 m/z ratios each) and for monoisotopic, +1, and +2 analyte profiles (7 m/z ratios each) in (b) and (c).&nbsp (b) partial profiles for the lock mass ion, m/z 415 ion, +1 and +2 profiles, and mass calibration ion.&nbsp (c) lock mass ion, m/z 430 ion, +1 and +2 profiles, and mass calibration ion.&nbsp The exact masses and relative abundances determined from the partial profiles for each analyte are given under each profile.&nbsp (d) calculated monoisotopic ion and +1 and +2 profiles for the composition, C24H31O2Br, at a resolving power of 20,000 (10% valley). The exact masses and relative abundances on the profiles are based on full, rather than partial profiles.
The contributions of the three +1 ions and the three most abundant +2 ions are shown.

Two chromatographic peaks arising from two isomers are apparent in the bottom three traces of
Figure 1a.&nbsp The top trace in Figure 1a, due to an ever-present lock mass ion, displays two baseline excursions that were induced by reversing the polarity of the draw-out plates in the ion source for 5 s to enable automated integration by the data system of the shaded area between the excursions.&nbsp The weighted average of the top several points across each partial profile provides its exact mass, while relative abundances are determined from ratios of the sums of points used to plot two profiles.&nbsp Partial profiles are also plotted using 5 m/z ratios each for two calibrant ions ("lock-mass" and "calibration mass" in Figure 1) from perfluorokerosene (PFK), with masses that bracket those of the analyte profile.&nbsp The mass errors for the two calibration ions are used to estimate and correct the errors for analyte profiles by linear interpolation.

Figure 1d shows the calculated full profiles corresponding to the analyte partial profiles plotted in
Figure 1c.&nbsp The monoisotopic profile results from a single ion, C24H31O2Br+, that contains only the most abundant isotope of each element (i.e., 12C, 1H, 16O, and 79Br).&nbsp The +1 mass peak profile arises from three contributing ions that contain a single atom of a +1 isotope, either 13C, 2H, or 17O.&nbsp The +2 profile is a composite from ions that contain a single atom of a +2 isotope, two atoms of the same +1 isotope, or one atom each of two +1 isotopes.&nbsp As shown in Figure 1d, the main contributors to the +1 and +2 profiles are the ions containing a single 13C or 81Br atom, respectively.&nbsp Only contributions from the three most abundant +2 ions are shown in the Figure.

MPPSIRD provides several important advantages over continuous scanning and single-point per profile SIR.&nbsp Acquisition cycles of 1 sec or less provide the speed to delineate chromatographic peaks, the high sensitivity of SIR is retained, a mass resolving power up to 20,000 (10% valley) is used routinely, and mass calibration stability is assured by recalibration against a lock mass at the start of each SIR cycle.&nbsp Major interferences or a lack thereof are documented by the shapes of the plotted profiles.&nbsp In general, if the shape of a plotted profile is Gaussian, mass interferences are minor or nonexistent.&nbsp Exceptions occur when coeluting compounds produce ions with different elemental compositions having almost identical exact masses.(9)

A three-step approach is used to determine ion compositions.&nbsp First, a survey of the mass range about an analyte ion mass is made at a resolving power of 3000 (10% valley), monitoring 21 m/z ratios using a 100 ppm increment.&nbsp An estimate of the analyte ion's exact mass is obtained and used as the centroid mass for the next experiment, viz. the plotting of a full profile with data acquired using a resolving power of 10,000 (10% valley) and a 10 ppm increment to provide 10 points across the profile.&nbsp Finally, with the same resolving power and increment, partial profiles are obtained using 7 m/z ratios each for the monoisotopic, +1, and +2 profiles.&nbsp From the three partial (7-point) profiles, three exact masses and two relative abundances are obtained.&nbsp Alternatively, to reduce error in relative abundances and to simplify explanations in legal proceedings, experiments can be performed to obtain the full (10-point) monoisotopic profile plus either the +1 or +2 full profile.(8) Full or partial profile data can be acquired for up to 32 different eluting compounds per injection.

Before monitoring partial profiles, a hypothetical composition must be chosen from the list of compositions that are possible based on the exact mass and its error limits for the monoisotopic profile.&nbsp Predicted masses for the monoisotopic, +1, and +2 partial profiles are calculated by a Profile Generation Model (PGM) and used to prepare the MID (i.e., SIR) descriptor.&nbsp The other compositions have different predicted masses.&nbsp Hence, the relative abundances measured using the predicted masses of the hypothetical composition would provide low values for other compositions, because their partial profiles would be off-center within the mass ranges monitored.&nbsp This point accounts for the low relative abundance ranges for non-hypothetical compositions in tables prepared by the PGM based on partial profiles.

For each experiment, a macro language program within a Lotus 123 version 9.0 spreadsheet (Lotus Development Corp., Cambridge, MA, USA) prepares ASCII files for execution by Finnigan MAT ICL 10.6 and ICIS 8.2.1 software (Thermo Finnigan, Bremen, Germany).&nbsp The spreadsheet resides on an ancillary Gateway G6-200 personal computer (Gateway Corp., Sioux City, SD, USA) operated by Windows NT version 4.0 software (Microsoft Corp., Bellevue, WA, USA).&nbsp The Finnigan software is part of the data system on the UNIX version 4.0 based Digital Alpha minicomputer (Digital Corp., Maynard, MA, USA).&nbsp After the user enters centroid masses of profiles from previous experiments or for a hypothetical composition, start and end times that bracket the chromatographic peak for each analyte-specific time window, the resolving power, and data file name, the ASCII files are prepared and sent to the data system computer by FTP across an ethernet connection.&nbsp One file prepares the instrument by autotuning, by entering 31 m/z ratios into the MID (SIR) view for each of up to 32 sets of profiles to be studied, and by adjusting the resolving power to the specified value.&nbsp During data acquisition, another file reverses the polarity of the draw plates to induce baseline excursions in the ever-present calibrant ion traces.&nbsp After data acquisition, a third file displays the ion chromatograms and saves m/z ratios, peak areas, and retention times in an ASCII file.&nbsp The original spreadsheet imports this report file using FTP and plots the profiles labeled with the measured exact masses and relative abundances.&nbsp Figures 1b and 1c show partial profiles for an m/z 415 fragment ion and its +1 and +2 profiles and for an m/z 430 molecular ion and its associated profiles.&nbsp A more detailed description of MPPSIRD is provided in References 10 and 11.&nbsp This software can be adapted for use with other double focusing mass spectrometers that provide a macro language for control of instrumental parameters and for display, processing, and ASCII file outputs of acquired data.

Profile Generation Model (PGM)

The experimental values are entered into a Profile Generation Model written in QuickBASIC version 4.50 (Microsoft Corp., Bellevue, WA, USA) residing on the personal computer.&nbsp The PGM determines the ion compositions that are possible based on the exact mass of the monoisotopic profile within 6 ppm when a resolving power of 10,000 (10% valley) is used and a single data acquisition is made.&nbsp The PGM then compiles a table of calculated exact masses, relative abundances, and relative abundance ranges that take into account sources of error associated with measuring either partial or full profiles.&nbsp An "X" is placed next to each calculated exact mass or relative abundance range that is inconsistent with the measured value, and the composition in that row is rejected.&nbsp Tables 1-6 were extracted from such PGM outputs.&nbsp The atomic masses and isotopic abundances used by the model are from Reference 12.&nbsp The PGM is described further in Reference 13.

Instrumentation

MPPSIRD was originally developed using a VG 70SE double focusing mass spectrometer (VG Analytical, Manchester, UK) and was later adapted to a Finnigan MAT 900S double focusing mass spectrometer (Thermo Finnigan, Bremen, Germany).&nbsp The exact mass and relative abundance error limits used in the PGM are based on data acquired with the VG instrument.(11,13) Although both instruments provided exact masses accurate to within 2 ppm for most measurements, uncertainty limits were chosen that exceeded the largest errors observed for 15 single measurements made with probe introduction for each of four standards.&nbsp Use of wide uncertainty limits reduces the chance of rejecting a correct composition due to the presence of minor interferences not detectable from visual inspection of ion chromatograms.&nbsp Cycle times were 0.8 sec with the VG instrument when monitoring 23 m/z ratios including only one calibrant ion, and 1.0 sec with the MAT instrument monitoring 31 m/z ratios.&nbsp Each m/z ratio was monitored for 30 msec with the VG or 20 msec with the MAT.&nbsp References 8-11, 13-17 provide specific experimental details for several applications of MPPSIRD and the PGM.

For the compound tentatively identified as a bromination product of Quinbolone, 1 µL of a methanol extract of chlorine-disinfected well water was injected onto a 30 m, 0.25 mm i.d., 25 µm coating, RTx -5 capillary column (Restek, Bellefonte, PA).&nbsp The temperature program was 100oC for 1 min, 10oC/min to 300oC, hold 1 min, and the He flow was 0.5 cc/min.&nbsp The transfer line was maintained at 260oC.&nbsp Electron impact ionization by 70 eV electrons occurred in an ion volume at 250oC.&nbsp The secondary electron multiplier voltage was 1.9 kV.&nbsp Each magnetic scan over 320 - 450 Da required 0.64 s and the resolving power was 800 (10% valley).&nbsp For the partial profiles obtained using MPPSIRD in Figures 1b and 1c, the resolving power was 20,000 (10% valley).

RESULTS AND DISCUSSION

SIR Acquisition Cycle Time

For extracts containing dozens of compounds, delineation of chromatographic peaks is essential for assigning fragment ions to the correct molecular ion.&nbsp Determination of compositions for ions from coeluting compounds would confound identification of an analyte.&nbsp To delineate gas chromatographic peaks, an SIR acquisition cycle of 1 s or less is required.

Exact mass and the number of possible compositions

To confirm the identities of synthetic or natural products based on ion compositions, the Journal of the American Society for Mass Spectrometry(18) specifies that the "acceptable uncertainty" in the exact mass measurement must be assessed, and all elemental compositions possible for an ion within that error range must be considered.&nbsp One cannot assume that the composition of a molecular ion has been unequivocally confirmed if its measured exact mass falls within some arbitrary error limit.&nbsp The number of possible compositions for an ion depends on its nominal mass and mass defect, the uncertainty limit of the mass determination, the maximum number of atoms of each element considered, and the range of rings plus double bonds deemed possible.

The elements considered depend on the source of the analyte.&nbsp For synthetic products, consideration of the elements available in the reaction vessel is sufficient,(9) but for complex environmental samples such as sewage treatment influents and effluents numerous elements are plausible.&nbsp If certain elements are excluded from consideration for a given ion, and a tentative identification is confirmed with a standard, omission of those elements becomes immaterial.&nbsp But if an element actually present in the ion is arbitrarily excluded, the true composition will not be found and the correct tentative identification will not be made.&nbsp Hence, it is best to include all elements that could conceivably be found in environmental contaminants.&nbsp The presence and number of atoms of Cl or Br are revealed by the high relative abundance of the mass peak with a m/z ratio greater than the monoisotopic mass peak by 2 Da in low resolution mass spectra, which results from the isotopic abundances of 37Cl (24.23%) or 81Br (49.31%).(12) The isotopic abundance of 29Si (4.67%) usually reveals the presence of multiple Si atoms in column bleed or septa components, or in silicones in the sample, by providing an elevated +1 profile.&nbsp For most organic contaminants, a +2 mass peak with an abundance between 4% and 32% relative to the monoisotopic profile indicates the presence of one or more S atoms.&nbsp However, poor accuracy for relative abundance measurements or interferences from ever-present or coeluting compounds might obscure the presence of a single S atom.

The numbers of atoms of C, H, N, O, F, and P in an ion are not usually discernible by interpretation of low resolution mass spectra.&nbsp Except for the arsenic containing compound discussed later, the elements C, H, N, O, F, P, and S were always considered to calculate the numbers of compositions discussed herein.&nbsp In practice, Cl, Br, or Si are also included if the isotopic pattern of an ion suggests their presence.&nbsp For example, the presence of a single Br atom in two ions was evident from the mass spectrum of the last compound to be discussed.

In Figure 2a, the numbers of possible compositions based on C, H, N, O, F, P, and S atoms as a function of nominal mass with 10, 5, and 2.5 ppm uncertainty limits, are plotted.

Figure 2 - For further information contact grange.andrew@epa.gov

Figure 2. (a) The number of possible compositions as a function of nominal masses with error limits of 10, 5, and 2.5 ppm considering C, H, N, O, F, P, and S atoms.&nbsp (b) the number of possible compositions for exact masses near 400 Da at 0.05 Da intervals based on C, H, N, O, F, P, and S atoms with up to 10 ppm error in the exact mass determination.&nbsp The numerals on or above the vertical bars are the maximum number of remaining possible compositions after considering three exact masses and two relative abundances obtained from a single set of partial profiles acquired with 20,000 resolving power (10% valley).

There are three points plotted for each nominal mass (e.g. 400.00000 ± 2.5 ppm [{rectangles} 470 compositions], ± 5 ppm [{diamonds} 932 compositions], and ± 10 ppm [{ovals} 1860 compositions]), with a mass increment between nominal masses of 10 Da&nbsp The number of possible compositions is approximately halved each time the error limit is halved.

In Figure 2b, the number of compositions as a function of the mass defect near a nominal mass of 400 Da is shown for exact masses between 399.50 and 400.50 Da at 0.05 Da intervals for ± 10 ppm error limits.&nbsp Each histogram bar corresponds to the number of possible compositions.&nbsp The number of possible compositions is greatest for m/z 399.95 Da and decreases toward lower or higher masses.

An elemental composition generator is part of the PGM and was used for these calculations.&nbsp To check the calculations, the numbers of possible compositions were calculated for 10 randomly chosen exact masses greater than 200 Da.&nbsp The numbers were identical to those calculated by the elemental composition generator included as part of a Finnigan MAT 900S data system.&nbsp The number of rings and double bonds was calculated as (x -1/2 y + 1/2 z + 1),(19) where x is the number of C atoms, y is the number of H and F atoms, and z is the number of N and P atoms.&nbsp While no maximum number of rings plus double bonds was specified, no fewer than -0.5 were permitted.

For a molecular mass of 400.00000 Da and an uncertainty limit of 5 ppm, 932 compositions are possible based the atomic masses of the seven elements considered.&nbsp As is evident from Figure 2b, use of ICE can reduce the number of possible compositions by more than an order of magnitude.&nbsp Still, multiple possible compositions would remain.&nbsp Answering the following questions could reject the majority of the remaining compositions.&nbsp Has any isomer of a given composition ever been synthesized?&nbsp Is the isomer chemically stable in the environment sampled?&nbsp Is the isomer on the list of high production volume chemicals, used in commerce, or possibly a byproduct or degradation product of such a compound?&nbsp Could the observed fragment ions arise from the isomer?&nbsp Have other analytical techniques provided structural details consistent with the isomer?

Other useful mass spectrometric data

Exact masses of +1 and +2 profiles

The exact mass differences between 13C and 12C (1.00335 Da), 2H and 1H (1.00628 Da), 15N and 14N (0.99703 Da), 17O and 16O (1.00422 Da), 18O and 16O (2.00425 Da), 33S and 32S (0.99939 Da) and 34S and 32S (1.99580 Da) yield unique exact masses for the +1 and +2 mass peak profiles for each possible composition.&nbsp Table 1 lists the compositions possible for 133.06400 Da with a 5 ppm uncertainty limit.

Table 1 - For further information contact grange.andrew@epa.gov

The number of atoms of each element considered is in the range from zero up to INT(133.06400/A) where INT provides the integer for the measured mass divided by each elemental mass.&nbsp The exact mass of the +1 profile would eliminate composition 1, while that of the +2 profile would reject all but the correct composition.&nbsp The +2 profile provides greater mass differences between compositions, but might not be measurable to within the desired uncertainty limits when produced from ultra-trace amounts of minor components in a complex mixture.&nbsp If exact mass measurements can be made for the +1 and +2 profiles, the number of possible compositions that must be addressed by chemical and commercial arguments is significantly reduced.

Relative abundances of +1 and +2 profiles

Relative abundances of the +1 and +2 profiles are additional mass spectral data that permit rejection of incorrect compositions when they can be measured with sufficient accuracy.&nbsp For low-mass ions, compositions can be determined from accurately measured relative abundances alone.(17) In Table 2, several compositions with similar exact masses for the monoisotopic, +1, and +2 profiles are rejected based on different relative abundances.

Table 2 - For further information contact grange.andrew@epa.gov

Accurate values of relative abundance for the +1 profile rejected those compositions with too few C atoms, while all compositions containing S atoms were among those inconsistent with the +2 values.

Linear dynamic range requirements for accurate relative abundance measurements

To make practical measurements of relative abundances of 1% or less for +2 profiles from ions not containing Cl, Br, S, or Si, a linear dynamic range of at least 3 orders of magnitude is needed to ensure both ion abundances fall within the linear response range.&nbsp If compounds in complex mixtures with very different concentrations are studied during the same data acquisition, a still wider linear dynamic range is important.

Interferences

For GC/MS analyses, numerous ions are generated from column bleed and provide potential interferences at many m/z ratios, especially as the column temperature is raised.&nbsp The severity of these interferences will depend upon whether exact masses for monoisotopic profiles or less abundant +1 or +2 profiles are being measured, and upon the mass difference between the analyte and interfering profiles.&nbsp In addition, numerous compounds coelute from complex environmental extracts.&nbsp Mass interferences often arise despite use of resolving powers of 10,000 and 20,000 (10% valley).

Manufacturers of double focusing mass spectrometers use the 10% valley definition of resolving power, while manufacturers of most other types of mass spectrometers use a Full Width at Half Maximum (FWHM) definition.&nbsp Figure 3a illustrates these two definitions.

Figure 3 - For further information contact grange.andrew@epa.gov

Figure 3. (a) Two hypothetical mass peak profiles of equal height, differing in mass by 50 ppm separated by a 10% valley, (b) profiles and experimentally determined exact masses of 3 ions having the same nominal mass observed with MPPSIRD at a resolving power of 5000 (10% valley), (c-d) calculated profiles for equal amounts of a column bleed and analyte ion for resolving powers of 3400 and 1700 (10% valley), and (e-f) calculated profiles for a hypothetical analyte ion closer in mass to the column bleed ion at the same resolving powers.

In this hypothetical example, two profiles with equal abundances overlap with a 10% valley between them.&nbsp The 10% valley resolving power is the average mass of the two profiles divided by the mass difference between the maxima (ave. m/delta m).&nbsp The Full Width at Half Maximum resolving power is the mass at the maximum of a single profile divided by the mass width across the profile at 50% of the maximum abundance (m/delta m50%).&nbsp The conversion factor between the two resolving power definitions is 2.082.&nbsp Hence, halving a FWHM resolving power approximates the corresponding 10% valley resolving power.&nbsp These definitions are mass independent.

Figure 3b shows three experimental mass peak profiles for m/z 281, due to a calibrant ion from PFK, a column bleed component, and an analyte.(15) The profiles were recorded using a resolving power of 5000 (10% valley).&nbsp These profiles are nearly baseline resolved and neither the calibrant nor column bleed ion contributed signal to the analyte ion.&nbsp This is not the case at lower resolving powers.&nbsp Figures 3c and 3d show calculated profiles at resolving powers of 1700 and 3400 (10% valley) for the same column bleed and analyte masses in Figure 3b, but with equal ion abundances.&nbsp Accurate exact masses could be assigned for both partially resolved profiles in Figure 3d, but not for the shifted maxima of the observed composite profile in Figure 3c.&nbsp Hence, for this analytical problem, only the higher resolving power (3400) would be adequate for measuring the analyte ion's exact mass.&nbsp Figures 3e and 3f illustrate a hypothetical example where this would not be true.&nbsp For both resolving powers, only a single maximum would be seen and an average exact mass would be obtained.&nbsp In this case, a resolving power of 10,000 (10% valley) would resolve the profiles and provide accurate exact masses for both profiles and an accurate relative abundance for the analyte profile.

Utility of oa-TOF mass spectrometers for determining ion compositions

A survey was made of nine recently published journal articles that used oa-TOF mass spectrometers coupled to GC,(20) LC,(4,21-26) or capillary zone electrophoresis(27) to measure exact masses of ions from pesticides or herbicides,(4,21,22) pharmaceuticals,(23- 25,27) petroleum,(20) or a protein digest.(26) Electrospray ionization was used for the LC studies and field-ionization(20) when GC was used.&nbsp These articles noted or implied three important advantages of oa-TOF MS.&nbsp First, spectral summation times for the thousands of data acquisitions made each second were between 0.2 s(23) and 2 s.(22) However, if a coelution question arose, a shorter integration time could be used to partially resolve chromatographic peaks of closely eluting compounds and to establish which fragment ions were formed from which molecular ions.&nbsp Second, after initial calibration, only one internal calibrant ion is required for the entire mass range, an advantage since a calibrant that provides few fragment ions provides few potential mass interferences.&nbsp Third, exact masses are obtainable for both the molecular ion and fragment ions from the same oa-TOF mass spectrum.

The maximum errors observed for single measurements were as high as 13 ppm using internal calibration(20) and 38 ppm using external calibration.(22) However, most measured exact masses fell within 10 ppm(4,20,21,27) and within 5 ppm(22-26) when at least three independent measurements were averaged.

Linear dynamic ranges of only 50,(27) 100,(4) and 200(26) were demonstrated in the oa-TOF articles.&nbsp In addition, the oa-TOF mass spectra in many figures did not provide accurate relative abundances.(4,23,27) Therefore, not even the +1 relative abundance can be used to estimate the number of C atoms in an ion.&nbsp Three groups used the +2 relative abundance obtained by oa-TOF MS to eliminate only Cl and Br atoms from consideration;(4,21,22) only one group used the oa-TOF data to also eliminate S atoms.(26)

The resolving powers used in the surveyed oa-TOF MS articles ranged from 3500 to 7000 (FWHM), which correspond to a range of 1700 to 3400 with the 10% valley definition.&nbsp Figures 3c through 3f illustrated that mass interferences from column bleed or coleuting compounds are possible at these resolving powers, although the use of only one mass calibrant ion removes potential mass interferences from PFK or a similar calibrant that provides numerous ions.

Utility of FTICR mass spectrometers for determining ion compositions

When FTICR mass spectrometry is coupled with chromatographic techniques, it is usually employed to study large biomolecules or pharmaceuticals.(28-33) For FTICR mass spectrometers, the stability of the ion cloud for each exact mass depends on the number of ions present.(29) This relationship can distort measurement of the relative abundances.

Although FTICR mass spectrometers are capable of resolving power much greater than 20,000 (10% valley) with long ion residence times in the trap, such resolving powers are not reported for the time scale of high resolution gas chromatography peaks.&nbsp To achieve 8000-11,000 (FWHM) resolving power, 4-s aquisition cycles were required using micro-ESI/LC FTICR.(29) Another group using external ion accumulation, an 8 T super cooled magnet, and five stages of differential pumping achieved 30,000 resolving power (FWHM), i.e., 14,400 resolving power (10% valley), with a data collection frequency of 2 Hz.(31) FTICR mass spectrometers can determine molecular ion compositions through MSn experiments based on fragmentation patterns, but relative abundances are only reported to have been used to indicate the presence of Cl or Br atoms.(34) Thus, in the present context, the main limitations of FTICR MS in its current state of development are the mass spectral acquisition time required to provide a sufficiently high resolving power and the insufficient accuracy of relative abundance measurements.

Utility of double-focusing mass spectrometers for determining ion compositions

Historically, double focusing mass spectrometers employing electronic data systems for detection, rather than the now rarely used photographic plates, could only measure exact masses through peak matching which required at least a minute, or full scanning with high mass resolving power which required relatively large amounts of analytes or several seconds per scan over a narrow mass range.(10) This speed limitation is still commonly stated to dismiss such instruments as a means for measuring exact masses of compounds directly introduced into a mass spectrometer after chromatographic separations.(20,22,27) Yet, for the past decade our laboratory has employed MPPSIRD (described in the experimental section) to routinely measure exact masses and relative abundances of ions using a resolving power of up to 20,000 (10% valley) with acquisition times sufficiently short to delineate chromatographic peaks as analytes elute from a gas chromatograph into a double focusing mass spectrometer.(8-11,13-17) Other groups have developed abbreviated scans over very narrow mass ranges (<1 Da)(35,36) or used selected ion recording to observe mass peak profiles as analytes eluted.(37) These groups documented the presence or absence of interferences and measured exact masses, but did not apply these techniques for determining unique ion compositions.&nbsp More recently, another group used MPPSIRD to determine the compositions of the molecular ion and fragment ions from a naturally occurring bioaccumlative organohalogen compound they found in seals, bird eggs, porpoises, and human milk.(38,39)

The numeral for each bar of the histogram in Figure 2b is the maximum number of possible compositions that could remain after three exact masses and two relative abundances from a single set of three partial profiles, acquired using a resolving power of 20,000 (10% valley), were entered into the PGM.&nbsp Multiple data acquisitions would reduce the error limits and reject all but the correct composition for each exact mass.(13) In practice, when fragment ions are observed, a single data acquisition is made for one or more fragment ions to accomplish the same end, while providing one or more additional ion compositions.

Currently, double focusing mass spectrometers using MPPSIRD provide the most accurate measurements of relative abundances(40) and provide a linear dynamic range of at least 104. Thus, double focusing mass spectrometers still provide the highest usable resolving power for high resolution GC/MS applications along with the accurate relative abundances so useful for determining ion compositions.

Real-world examples

A low-mass well contaminant

Using ICE with a double focusing instrument, several isomeric compounds were identified in an extract of municipal well water sampled near Toms River, NJ, where increased incidences of childhood cancer and leukemia had been observed.(16) This section considers the question as to whether this identification could have been made using oa-TOF MS.

Listing possible compositions for fragment ions reduces the number of possible compositions for the molecular ion.&nbsp As shown in Figure 2a, an ion with a mass of 400.00000 Da ± 5 ppm has 932 possible compositions.&nbsp For comparison, only 29 compositions can correspond to a fragment ion or composite neutral loss of 200.00000 Da ± 5ppm.&nbsp In addition, the possible fragments of the molecular ion must be sub-units of possible molecular ion compositions.

Assuming that the exact masses of the molecular and fragment ions could be measured to within 5 ppm (as would be the case for an average of three or more measurements made with an oa-TOF MS), Table 3 was created to determine which molecular ions could produce the observed fragment ions.

Table 3 - For further information contact grange.andrew@epa.gov

Consideration of the five most abundant fragment ions rejected only 4 of 10 possible molecular ion compositions (compositions 1, 4, 5, and 7).&nbsp If similarly reliable exact mass values for the +1 profiles were obtained, composition A for the m/z 115 ion would be rejected.&nbsp The two remaining m/z 115 fragments could not be produced by molecular ion compositions 2 or 6; hence, these two compositions would also be eliminated.&nbsp Similarly, molecular ion composition 3 could be rejected based on rejection of any one of the compositions of the corresponding fragment ions at m/z 140, 153, and 156.&nbsp If the exact masses of the +2 profiles were also determined to within 5 ppm, molecular ion composition 8 could be eliminated due to rejection of any one of the corresponding fragment ions for all but the m/z 115 fragment ion.&nbsp Compositions 9 and 10, with exact masses within 1.5 ppm of each other, would remain viable.

For comparison, Table 4 shows a PGM output based on entry of the three calculated masses and two relative abundances for the mass peak profiles of C14H14N2.

Table 4 - For further information contact grange.andrew@epa.gov

The model predicts the relative abundances would reject all but the correct composition.&nbsp As expected, experimental determination of exact masses and relative abundances provided unequivocal evidence for this composition.(16) Using C14H14N2 to establish elemental limits of 14 C, 14 H, 2 N and no atoms of other elements, full mass peak profiles obtained at 10,000 resolving power (10% valley) provided a unique composition for each fragment ion.(16) In addition, the exact masses of the composite neutral losses were determined by subtracting the exact masses of fragment ions from that of the molecular ion, and their compositions were also determined using the PGM.(16)

If an oa-TOF MS were used to acquire three data acquisitions to achieve 5 ppm accuracy, the analyses would provide between two and six possible molecular ion compositions, depending on whether or not accurate exact masses could be obtained for the +1 and +2 profiles.&nbsp The ICE determination also required three injections, one each for three different MID descriptors.&nbsp By contrast, however, ICE provided the correct composition and more compelling evidence for correlating the isomeric compounds with their source (8) than would be provided by existing oa-TOF MS.

A compound containing arsenic

The mass spectrum for a trace-level compound in an extract of water from a monitoring well at a landfill displayed ions at m/z 182 and 167.&nbsp Other fragment ions were obscured by the chemical noise.&nbsp The compound was hypothesized to be 2-methyl-1,3,2-dithiarsolane and a standard was synthesized.&nbsp The standard and trace-level compound provided the same retention time and mass spectrum.&nbsp ICE was used to confirm that the composition of the molecular ion from the trace-level compound was C3H7S2As.&nbsp All three exact masses and both relative abundances were consistent with only this composition, as illustrated in Table 5 for data acquired using 10,000 resolving power (10% valley) and a corresponding error limit of 6 ppm.

Table 5 - For further information contact grange.andrew@epa.gov

Assuming oa-TOF MS provided an error limit of ± 5 ppm, eight of the 11 compositions would be possible about the calculated exact mass of 181.92051 Da.&nbsp If the exact mass of the +1 profile could also be measured by oa-TOF MS to within 5 ppm, four compositions would remain viable, and a similarly accurate measurement of the +2 profile's exact mass would leave three possible compositions.&nbsp The measured +2 relative abundance of 8.54% for a partial profile provided by MPPSIRD selected composition 8, the only composition with two S atoms, as the correct one.

It is instructional to consider how the outcome of this example would change if the presence of arsenic had not been known.&nbsp Based on consideration of only C, H, N, O, F, P, and S atoms and a 5 ppm error limit, four compositions would be possible: H6O3S4, HNOF3P2S, CHN2O3P3, and C3HNOFS3.&nbsp While none of these compositions could lose a methyl group, three could lose NH to account for the m/z 167 ion.&nbsp If the exact masses within 5 ppm of the correct value could be obtained for the +1 or +2 profiles with oa-TOF MS, all four compositions would be rejected.&nbsp Whereas this would suggest the presence of an unconsidered element, it would provide no clue to its identity.

The experimental values obtained using MPPSIRD with a 6 ppm error limit provided seven possible compositions.&nbsp All seven would be rejected based on the exact masses of the +1 and +2 profiles or on the +2 relative abundance, since none contained two S atoms.&nbsp Monitoring of full profiles (8) provides exact masses and relative abundances without having to select a hypothetical composition to provide centroid masses before acquiring data to plot partial profiles.&nbsp Full profiles would provide relative abundances within the ranges 4.26-5.72% for %+1 and 7.79-10.13% for %+2.&nbsp Two S atoms contribute 8.86% to %+2.&nbsp Hence, two S atoms in the ion would be indicated.&nbsp The two S atoms would contribute 1.58% to %+1, leaving (2.68-4.14)% to be accounted for, primarily by C atoms.&nbsp Contributions to %+1 from two, three, or four C atoms would be 2.20%, 3.30%, and 4.40%, respectively.&nbsp Three C atoms best fit the remaining range for %+1.&nbsp Two S atoms and three C atoms account for 100 Da, leaving 82 Da.&nbsp The remainder of the molecule will likely contain several H atoms and atoms of one or more monoisotopic elements.&nbsp P atoms were already considered and not present in any possible compositions.&nbsp Iodine atoms with a mass of 127 Da are too heavy.&nbsp Mn (55 Da) and Co (59 Da) atoms are too light.&nbsp But a single As atom with a mass of 75 Da is plausible.&nbsp Subtracting 175 Da from 182 Da would leave seven H atoms and suggest C3H7S2As as the ion composition.&nbsp This hypothesis would then be confirmed experimentally using ICE.

In this example, if the presence of As had not been assumed beforehand, oa-TOF MS would have provided four incorrect compositions or suggested that not all necessary elements were considered based on the exact masses of the +1 and +2 profiles, while the relative abundances measured using MPPSIRD would have led deductively to the overlooked element and the correct composition.

A high-mass disinfection byproduct

The low resolution mass spectrum shown in Figure 4a, with two mass peaks at m/z 415 and 417 visible above the chemical noise, suggested a mono-brominated compound might be present in a chlorine-disinfected, well-water extract provided by an EPA regional lab.

Figure 4 - For further information contact grange.andrew@epa.gov

Figure 4.&nbsp (a) A raw low resolution mass spectrum for a disinfection byproduct in a well water extract and (b) the background subtracted mass spectrum with the structure of a candidate compound.

Related ions with m/z 430 and 432 became apparent in the background subtracted mass spectrum in Figure 4b for which no library matches were found.&nbsp No other fragment ions were discernable above the chemical noise based on visual inspection of ion chromatograms.&nbsp The exact mass of the apparent molecular ion was determined to be 430.15123 Da.&nbsp Assuming the presence of a single Br atom, this exact mass with an error limit of 5 ppm corresponds to only 41 possible compositions.&nbsp Subtraction of the Br mass yields 351.23241 Da for the remainder of the ion.&nbsp As illustrated by Figure 2b, this mass defect results in relatively few possible compositions.&nbsp Only one composition could be eliminated if the exact mass of the +1 profile were known.&nbsp The exact mass of the +2 profile would eliminate no compositions, because the atom of 81Br is primarily responsible for the abundance and exact mass difference of the +2 profile relative to the monoisotopic profile in all compositions.&nbsp At 5 ppm mass accuracy, the current practical limit for oa-TOF MS, there would remain at least 40 possible compositions for this ion.

Figures 1b and 1c show the partial profiles obtained for the m/z 415 and 430 ions and the profiles arising from ions containing higher-mass isotopes acquired with 20,000 resolving power (10% valley) to reduce the error limit to 3 ppm.&nbsp Table 6 provides the last seven possible compositions listed by the PGM for both the apparent molecular ion and the fragment ion.

Table 6 - For further information contact grange.andrew@epa.gov

For the list of possible compositions based on the exact masses of these ions, the relative abundance of the +1 profile arising primarily from 13C atoms rejected all but the correct composition in both cases.

The exact mass of the neutral loss was measured as 430.15123 - 415.12744 = 15.02379 Da, which, as expected, corresponded to a methyl group (15.02348 Da).&nbsp ICE employed weighted averages of exact masses of ions, summation of their isotopic abundances, and consideration of the valences of elements to reject all but the correct composition.&nbsp Only when considering possible isomers of this single composition must chemical and commercial factors be invoked to make a tentative identification.

Chlorination of the well water containing bromide ions could result in bromination of organic compounds.&nbsp The structure of Quinbolone, an anabolic steroid, is shown in Figure 4b.&nbsp Three possible allylic bromination sites can account for the two isomers observed in the ion chromatograms in Figure 1a.&nbsp Substitution of a Br atom for an H atom would provide the observed composition.&nbsp This is of interest since a feed lot was located near the well and anabolic steroids are often used to stimulate growth.&nbsp Purchase of Quinbolone, its chlorination in the presence of bromide ions, and examination of the mass spectra of the products would be logical next steps in the identification process for this compound.

In this case oa-TOF MS with 5 ppm uncertainty could provide, at best, a list of 40 possible compositions, while ICE performed on a double focusing mass spectrometer provided only one.&nbsp This information was sufficient to hypothesize the identity of the compound and to plan a strategy for confirming or rejecting this very tentative identification based on a practical chemical and commercial literature search.

Suggestions for further work

These examples suggest that a bench-top double focusing mass spectrometer, closer in price to an oa-TOF MS than the instruments currently used to perform ICE, might be sufficiently powerful for determining ion compositions for compounds eluting from a GC.&nbsp The wide linear dynamic range anticipated for one such commercially available instrument that has similar resolving power to oa-TOF instruments should provide relative abundances sufficiently accurate to reject many possible ion compositions that remain viable after consideration of exact masses alone.&nbsp In addition, such an instrument might provide more convenient quantification of compounds with dissimilar concentrations.

Manufacturers of oa-TOF instruments are continually improving the capabilities of their instruments.&nbsp If linear dynamic ranges in excess of 103 become available, and if these instruments are then capable of making more accurate relative abundance measurements for the +1 and +2 profiles, they will become far more useful for determining ion compositions.&nbsp Improvements in the resolving power would also be helpful when mass interferences are a problem.&nbsp Such an oa-TOF mass spectrometer could be used to limit the possible compositions for many prominent ions in a mass spectrum based on three or fewer data acquisitions.&nbsp Thereafter, only a few ions might need to be investigated using ICE to establish which compositions were correct for each ion.&nbsp Fewer experiments would be necessary to determine the compositions of all major ions in the mass spectrum than by using ICE alone.&nbsp Hence, oa-TOF and double focusing mass spectrometers could complement each other for identifying or characterizing the myriad of compounds present at ultra-trace to higher levels in many environmental samples, but for which library mass spectra are not available.

CONCLUSION

The utility of a mass spectrometer for determining ion compositions is limited by its data acquisition speed, mass accuracy, linear dynamic range, and often by its resolving power.&nbsp Currently, double focusing mass spectrometers equipped to perform ICE provide the most advantageous combination of these capabilities.

The oa-TOF instruments described in the literature are inexpensive relative to full-size double focusing instruments, but are not as powerful for determining ion compositions.&nbsp Absent interferences, averaged oa-TOF mass spectra often provide exact masses of ions to within 5 ppm.&nbsp This mass accuracy is useful for obtaining an exact mass consistent with a particular composition, but is not usually sufficient for proving conclusively that the ion has that composition.&nbsp For ions with masses less than 150 Da, fewer than 10 possible compositions are possible based on consideration of C, H, N, O, F, P, and S atoms.&nbsp Most of these compositions can be rejected based on consideration of any fragment ions observed, chemical knowledge, and chemical production levels.&nbsp For unidentified compounds, only a few compositions remain and literature searches for the compound would be feasible.&nbsp However, for higher mass ions with hundreds of possible compositions, the relative abundance measurements provided by ICE using a double focusing mass spectrometer are essential for selecting unique compositions for the molecular ion and fragment ions.

FTICR MS provides unequaled resolving power and mass accuracy as long as sufficient time is available to acquire the necessary degree of information in the time-domain.&nbsp The problem in the present context is that this required time is too long to allow sufficient spectra to be obtained across a high-resolution GC peak for the peak to be properly delineated.&nbsp In addition, accurate relative abundances are not obtainable.

A compound's identity is usually verified by comparison of its mass spectrum and retention time to those of a standard.&nbsp The correct tentative identification leading to purchase or synthesis of the standard can be reached by using chemical intuition to discard many unlikely compositions, searching multiple data bases, making extensive literature searches, and making limiting assumptions concerning the elements present.&nbsp But greater efficiency is realized by using ICE to mechanically determine the composition of the molecular and fragment ions through automated data acquisition and interpretation based on addition of atomic masses and naturally occurring isotopic abundances and on elemental valence restraints.&nbsp These ion compositions greatly limit the possible compounds to a number of isomers before chemical expertise is invoked.&nbsp Hence, the interpretive effort of the analyst and the number of standards purchased before a compound identity is confirmed will generally be less when a double focusing mass spectrometer with ICE is used rather than an instrument providing lower resolving power and relative abundances with accuracy only sufficient to reveal Cl and Br atoms.&nbsp Because ICE determines ion compositions based on physical properties, rather than chemical speculation, it is more likely to lead to compound identities than studies employing analytical techniques less powerful for this task.&nbsp For legal proceedings, tables of unique compositions for the ions in the mass spectrum and the composite neutral losses that produced the fragment ions provide a preponderance of evidence for the compound's identity unlikely to be challenged. (9,16) When an environmental contaminant cannot be identified, the ion compositions, along with a retention time, can unequivocally track it back to its source. (8)

ICE is an important advance in compound identification, but cannot by itself confirm the identities of compounds, since multiple isomers are usually possible.&nbsp Indeed, in our studies of complex environmental extracts, dozens of ion compositions for unidentified compounds remain.&nbsp Yet, for many analytes in environmental samples, ICE provides the best chance for compound identification.&nbsp For compounds found in extracts showing toxicological effects that remain unidentified after determination of ion compositions, NMR studies on synthetic standards and larger amounts of the compounds isolated from the extract would be warranted.(41)

REFERENCES

  1. Colborn T, Dumanoski D, Myers JP Our Stolen Future Penguin:&nbsp New York, 1996.
  2. Osemwengie LI, Steinberg S J. of Chromatog. A 2001; 932: 107-118.
  3. Snyder SA, Kelly KL, Grange AH, Sovocool GW, Snyder EM, Giesy JP Pharmaceuticals and Personal Care Products in the Environment:&nbsp Scientific and Regulatory Issues; Daughton CG, and Jones-Lepp T, Eds.; Symposium Series 791; American Chemical Society: Washington, DC, 2001; 116-141.
  4. Hogenboom AC, Niessen WMA, Little D, Brinkman UATh Rapid Commun. Mass Spectrom.1999; 13: 125-133.
  5. http://www.epa.gov/opptintr/chemrtk/hpv_1990.pdf [PDF, 205 pp., size not available, About PDF]
  6. Hau J, Stadler R, Jenny TA, Fay LB Rapid Commun. Mass Spectrom. 2001; 15: 1840-1848.
  7. Grange AH, Osemwengie L, Brilis G, Sovocool GW Internat. J. of Environmental Forensics 2001; 2: 61-74.
  8. Grange AH, Sovocool GW J. AOAC Internat. 1999; 82: 1443-1457.
  9. Grange AH, Donnelly JR, Brumley WC, Billets S., Sovocool GW Anal. Chem. 1994; 66: 4416-4421.
  10. Grange AH, Donnelly JR, Sovocool GW, Brumley WC Anal. Chem. 1996; 68: 553-560.
  11. CRC Handbook of Chemistry and Physics 74th ed., Section 1, pp. 10-11, Lide, D.R. editor, Boca Raton, FL, CRC Press, 1994.
  12. Grange AH, Brumley WC J. Amer. Soc. Mass Spectrom. 1997; 8: 170-182.
  13. Grange AH, Brumley WC Rapid Comm. in Mass Spectrom. 1992; 6: 68-70.
  14. Grange AH, Brumley WC LC.GC 1996; 14: 978-986.
  15. Grange AH, Sovocool GW, Donnelly JR, Genicola FA, Gurka DF Rapid Commun. Mass Spectrom. 1998; 12: 1161-1169.
  16. Grange AH, Sovocool GW Rapid Commun. Mass Spectrom. 1999; 13: 673-686.
  17. Instructions for Authors J. Am. Soc. Mass Spectrom. 2002; 13: A24.
  18. McLafferty, F.W., & Turcek, F. (1993) Interpretation of Mass Spectra, 4th Ed., Univ. Science Books, Mill Valley, CA, P. 28.
  19. Hsu CS, Green M Rapid Commun. Mass Spectrom. 2001; 15: 236-239.
  20. Bobeldijk I, Vissers JPC, Kearney G, Major H, van Leerdam JA J. Chromatog. A 2001; 929:63-74.
  21. Maizels M, Budde WL Anal. Chem. 2001; 73: 5436-5440.
  22. Jiang L, Moini M Anal. Chem. 2000; 72: 20-24.
  23. Wolff JC, Eckers C, Sage AB, Giles K, Bateman R Anal. Chem. 2001; 73: 2605-2612.
  24. Eckers C, Wolff JC, Haskins NJ, Sage AB, Giles K, Bateman R Anal. Chem. 2000; 72: 3683-3688.
  25. Zhang H, Heinig K, Henion J J. Mass Spectrom. 2000; 35: 423- 431.
  26. Palmer ME, Clench MR, Tetler LW, Little DR Rapid Commun. Mass Spectrom. 1999; 13: 256-263.
  27. Pinto DM, Boyd RK, Volmer Da Anal. Bioanal. Chem. 2002; 373: 378-389.
  28. Bresson JA, Anderson GA, Bruce JE, Smith RD J. Am. Soc. Mass Spectrom. 1998; 9: 799-804
  29. Emmett MR, White FM, Hendrickson CL, Shi SDH, Marshall AG J. Am. Soc. Mass Spectrom. 1998; 9: 333- 340.
  30. Winger BE, Kemp CAJ Am. Pharmaceutical Rev. 2001; 4: 55-63
  31. Li W, Hendrickson CL, Emmett MR, Marshall AG Anal. Chem. 1999; 71: 4397-4402.
  32. Lingjun L, Masselon CD, Anderson GA, Pasa-Tolic L, Lee SW, Shen Y, Zhao R, Lipton MS, Conrads TP, Tolic N, Smith RD Anal. Chem. 2001; 73: 3312-3322.
  33. Wu Q Anal. Chem. 1998; 70: 865-872.
  34. Tondeur Y, Hass JR, Harvan DJ, Albro PW Anal. Chem. 1984; 56: 373-376.
  35. Tong HY, Giblin DE, Lapp, RL, Monson S. J, Gross ML Anal. Chem. 1991; 63: 1772-1780.
  36. Alan AR, Roboz J Rapid Commun. Mass Spectrom. 1988; 2: 246- 249.
  37. Vetter W, Alder L, Palavinskas R. Rapid Commun. Mass Spectrom., 1999; 13: 2118-1224.
  38. Wu J, Vetter W, Gribble GW, Schneekloth JS Jr, Blank DH, Gorls H, Angewandte Chemie International Edition 2002; 41: 1740-1743.
  39. Burlingame AL, Boyd RK, Gaskell SJ Anal. Chem. 1998; 70: 647R- 716R.
  40. Nukaya H, Shiozawa T, Tada A, Terao Y, Ohe T, Watanabe T, Asanoma M, Sawanishi H, Katsuhara T, Sugimura T, Wakabayashi K Genetic Toxicology and Environmental Mutagenesis 2001; 492: 73-80.

Notice: The U.S. Environmental Protection Agency (EPA), through its Office of Research and Development (ORD), funded and performed the research described. This manuscript has been subjected to the EPA's peer and administrative review and has been approved for publication.

 

Analytical Environmental Chemistry
ICE Home Page

Environmental Sciences | Office of Research & Development
 National Exposure Research Laboratory
Author: Andrew Grange
Email: grange.andrew@epa.gov


Local Navigation


Jump to main content.